Call operations research
Auditing how often a caller's account changes during a call
An audit method for mid-call account changes, covering field-level coding, recheck rules, consent, and a review trail.

Table of contents
A caller can change a date, a name, or a contact route in the middle of a call. The change may be a correction, a new preference, or a different request altogether. An audit of those changes shows how often the first capture was wrong. It also shows whether the agent rechecked the revised facts before saving them.
Key takeaways
- Count changes by field, not only by call.
- Separate a correction from a new request.
- Check whether the revised value was read back.
- Record whether consent changed with the field.
- Use the pattern to fix forms and scripts, not to judge callers.
Research question and scope
The question is how often a caller's account changes during a call and which fields change most. The unit is the changed field within a call. A call with three changed fields counts as three field changes. The call itself counts once in the call-level rate.
The scope covers inbound calls that update an existing record or complete a new intake. It includes changes made before and after the record is saved. It excludes changes made by staff after the call without the caller present. Those changes are a separate control.
Record the form version, the queue, and the period. A form change can alter which fields are editable. The reader needs to know the version that produced the result. The partial intake review uses a similar version record.
Definitions and coding
Define a change as a value that differs from the value captured earlier in the same call. A typo correction counts. A change from an old address to a new address also counts. The definition should not depend on the agent's intent.
Define a correction as a change to a fact the caller already stated. Define a new request as a change that adds a different service, date, or purpose. The two are different for training. A correction points to a capture problem, while a new request points to a conversation that evolved.
Define a recheck as a readback or a confirmation of the revised value. A readback of the whole record counts only if the changed field is included. Record the method used, such as a spoken readback or a written confirmation. The call note template shows the operational format.
Sampling and audit steps
Select calls from a fixed period and a fixed queue. Include every call with a recorded change if the volume is small. Draw a random sample if the volume is large. Keep the filter and the export date in the report.
Read the note from top to bottom before coding. Mark the first value and the later value for each changed field. Use the timestamps where they exist. A note without timestamps is harder to code, and that difficulty should be reported.
Use two coders for a subsample. Compare their field-level codes before coding the full sample. A disagreement often reveals an ambiguous definition. Resolve it in writing so the next audit can use the same rule. The data reliability guidance supports this step.
Consent and sensitive fields
Check whether the change affects consent. A new phone number may need a new agreement for calls. A new email may need a new agreement for written messages. A change in purpose may need a fresh consent question.
Do not copy sensitive values into the audit file. Use a masked value or a reference. The audit needs to know that the field changed, not the full new value. This keeps the review inside the office's access rules.
Treat an unknown consent state as absent for follow-up. A changed channel without a clear agreement should not be used for outreach. The contact preference study offers a related coding frame. A conservative rule protects the caller and the office.
Measures and reporting
Report the number of calls with at least one change and the share of all calls. Then report the field-level counts. Show the counts beside the shares. A field that changes often in a small sample may not be a stable pattern.
Report the share of changes that were rechecked. Group the result by field and by agent. A low recheck rate on the phone number is a different risk from a low rate on a note field. The grouping makes the result actionable.
Report the outcome of the change when it is known. Did the caller confirm the new value, decline it, or leave the call? An unconfirmed change is a pending risk. The email readback study shows how capture errors can survive a weak check.
Limitations
The audit cannot show whether the first value or the second value was correct. It only shows that the value changed. A caller may correct a mistake, or may introduce one. The recheck record is the best evidence of which value the caller accepted.
The note may omit changes that happened verbally. An agent may update the record without writing a change block. Those changes are invisible to the audit. Report the limitation and improve the template before drawing strong conclusions.
The sample may overrepresent complex calls. Calls with many changes are more likely to be reviewed. That bias can inflate the change rate. Use a random sample and report the selection method. The abandoned call review discusses a similar selection issue.
Repeating the audit
Repeat the audit after a template change with the same definitions. A new change block may make more changes visible without changing caller behavior. Record the template version and the dates. The comparison should separate visibility from behavior.
Share the result with the front desk. Ask which fields are hardest to capture on the first pass. Their answers may point to a form layout or a script question. Then change one step and audit again.
The aim is a record that stays accurate through the conversation. A change is not a failure if it is captured and checked. The audit shows where the process is strong and where it needs a better question. One checked change is better than a silent rewrite.
Sources
- U.S. Government Accountability Office, GAO-20-283G: assessing data reliability
- National Institute of Standards and Technology, cybersecurity framework
- U.S. Department of Health and Human Services, HIPAA minimum necessary requirement
- International Organization for Standardization, ISO 18295-1 customer contact centres